linux.git/mm/memcontrol.c, branch v5.1

mm: writeback: use exact memcg dirty counts

2019-04-06T02:02:31+00:00

Since commit a983b5ebee57 ("mm: memcontrol: fix excessive complexity in
memory.stat reporting") memcg dirty and writeback counters are managed
as:

 1) per-memcg per-cpu values in range of [-32..32]

 2) per-memcg atomic counter

When a per-cpu counter cannot fit in [-32..32] it's flushed to the
atomic.  Stat readers only check the atomic.  Thus readers such as
balance_dirty_pages() may see a nontrivial error margin: 32 pages per
cpu.

Assuming 100 cpus:
   4k x86 page_size:  13 MiB error per memcg
  64k ppc page_size: 200 MiB error per memcg

Considering that dirty+writeback are used together for some decisions the
errors double.

This inaccuracy can lead to undeserved oom kills.  One nasty case is
when all per-cpu counters hold positive values offsetting an atomic
negative value (i.e.  per_cpu[*]=32, atomic=n_cpu*-32).
balance_dirty_pages() only consults the atomic and does not consider
throttling the next n_cpu*32 dirty pages.  If the file_lru is in the
13..200 MiB range then there's absolutely no dirty throttling, which
burdens vmscan with only dirty+writeback pages thus resorting to oom
kill.

It could be argued that tiny containers are not supported, but it's more
subtle.  It's the amount the space available for file lru that matters.
If a container has memory.max-200MiB of non reclaimable memory, then it
will also suffer such oom kills on a 100 cpu machine.

The following test reliably ooms without this patch.  This patch avoids
oom kills.

  $ cat test
  mount -t cgroup2 none /dev/cgroup
  cd /dev/cgroup
  echo +io +memory > cgroup.subtree_control
  mkdir test
  cd test
  echo 10M > memory.max
  (echo $BASHPID > cgroup.procs && exec /memcg-writeback-stress /foo)
  (echo $BASHPID > cgroup.procs && exec dd if=/dev/zero of=/foo bs=2M count=100)

  $ cat memcg-writeback-stress.c
  /*
   * Dirty pages from all but one cpu.
   * Clean pages from the non dirtying cpu.
   * This is to stress per cpu counter imbalance.
   * On a 100 cpu machine:
   * - per memcg per cpu dirty count is 32 pages for each of 99 cpus
   * - per memcg atomic is -99*32 pages
   * - thus the complete dirty limit: sum of all counters 0
   * - balance_dirty_pages() only sees atomic count -99*32 pages, which
   *   it max()s to 0.
   * - So a workload can dirty -99*32 pages before balance_dirty_pages()
   *   cares.
   */
  #define _GNU_SOURCE
  #include 
  #include 
  #include 
  #include 
  #include 
  #include 
  #include 
  #include 
  #include 

  static char *buf;
  static int bufSize;

  static void set_affinity(int cpu)
  {
  	cpu_set_t affinity;

  	CPU_ZERO(&affinity);
  	CPU_SET(cpu, &affinity);
  	if (sched_setaffinity(0, sizeof(affinity), &affinity))
  		err(1, "sched_setaffinity");
  }

  static void dirty_on(int output_fd, int cpu)
  {
  	int i, wrote;

  	set_affinity(cpu);
  	for (i = 0; i < 32; i++) {
  		for (wrote = 0; wrote < bufSize; ) {
  			int ret = write(output_fd, buf+wrote, bufSize-wrote);
  			if (ret == -1)
  				err(1, "write");
  			wrote += ret;
  		}
  	}
  }

  int main(int argc, char **argv)
  {
  	int cpu, flush_cpu = 1, output_fd;
  	const char *output;

  	if (argc != 2)
  		errx(1, "usage: output_file");

  	output = argv[1];
  	bufSize = getpagesize();
  	buf = malloc(getpagesize());
  	if (buf == NULL)
  		errx(1, "malloc failed");

  	output_fd = open(output, O_CREAT|O_RDWR);
  	if (output_fd == -1)
  		err(1, "open(%s)", output);

  	for (cpu = 0; cpu < get_nprocs(); cpu++) {
  		if (cpu != flush_cpu)
  			dirty_on(output_fd, cpu);
  	}

  	set_affinity(flush_cpu);
  	if (fsync(output_fd))
  		err(1, "fsync(%s)", output);
  	if (close(output_fd))
  		err(1, "close(%s)", output);
  	free(buf);
  }

Make balance_dirty_pages() and wb_over_bg_thresh() work harder to
collect exact per memcg counters.  This avoids the aforementioned oom
kills.

This does not affect the overhead of memory.stat, which still reads the
single atomic counter.

Why not use percpu_counter? memcg already handles cpus going offline, so
no need for that overhead from percpu_counter.  And the percpu_counter
spinlocks are more heavyweight than is required.

It probably also makes sense to use exact dirty and writeback counters
in memcg oom reports.  But that is saved for later.

Link: http://lkml.kernel.org/r/20190329174609.164344-1-gthelen@google.com
Signed-off-by: Greg Thelen 
Reviewed-by: Roman Gushchin 
Acked-by: Johannes Weiner 
Cc: Michal Hocko 
Cc: Vladimir Davydov 
Cc: Tejun Heo 
Cc: 	[4.16+]
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm/memcontrol.c: fix bad line in comment

2019-03-06T05:07:21+00:00

Commit 230671533d64 ("mm: memory.low hierarchical behavior") missed an
asterisk in one of the comments.

  mm/memcontrol.c:5774: warning: bad line:                | 0, otherwise.

Link: http://lkml.kernel.org/r/20190301143734.94393-1-cai@lca.pw
Acked-by: Souptick Joarder 
Signed-off-by: Qian Cai 
Reviewed-by: Andrew Morton 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm: remove zone_lru_lock() function, access ->lru_lock directly

2019-03-06T05:07:21+00:00

We have common pattern to access lru_lock from a page pointer:
	zone_lru_lock(page_zone(page))

Which is silly, because it unfolds to this:
	&NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]->zone_pgdat->lru_lock
while we can simply do
	&NODE_DATA(page_to_nid(page))->lru_lock

Remove zone_lru_lock() function, since it's only complicate things.  Use
'page_pgdat(page)->lru_lock' pattern instead.

[aryabinin@virtuozzo.com: a slightly better version of __split_huge_page()]
  Link: http://lkml.kernel.org/r/20190301121651.7741-1-aryabinin@virtuozzo.com
Link: http://lkml.kernel.org/r/20190228083329.31892-2-aryabinin@virtuozzo.com
Signed-off-by: Andrey Ryabinin 
Acked-by: Vlastimil Babka 
Acked-by: Mel Gorman 
Cc: Johannes Weiner 
Cc: Michal Hocko 
Cc: Rik van Riel 
Cc: William Kucharski 
Cc: John Hubbard 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

numa: make "nr_node_ids" unsigned int

2019-03-06T05:07:19+00:00

Number of NUMA nodes can't be negative.

This saves a few bytes on x86_64:

	add/remove: 0/0 grow/shrink: 4/21 up/down: 27/-265 (-238)
	Function                                     old     new   delta
	hv_synic_alloc.cold                           88     110     +22
	prealloc_shrinker                            260     262      +2
	bootstrap                                    249     251      +2
	sched_init_numa                             1566    1567      +1
	show_slab_objects                            778     777      -1
	s_show                                      1201    1200      -1
	kmem_cache_init                              346     345      -1
	__alloc_workqueue_key                       1146    1145      -1
	mem_cgroup_css_alloc                        1614    1612      -2
	__do_sys_swapon                             4702    4699      -3
	__list_lru_init                              655     651      -4
	nic_probe                                   2379    2374      -5
	store_user_store                             118     111      -7
	red_zone_store                               106      99      -7
	poison_store                                 106      99      -7
	wq_numa_init                                 348     338     -10
	__kmem_cache_empty                            75      65     -10
	task_numa_free                               186     173     -13
	merge_across_nodes_store                     351     336     -15
	irq_create_affinity_masks                   1261    1246     -15
	do_numa_crng_init                            343     321     -22
	task_numa_fault                             4760    4737     -23
	swapfile_init                                179     156     -23
	hv_synic_alloc                               536     492     -44
	apply_wqattrs_prepare                        746     695     -51

Link: http://lkml.kernel.org/r/20190201223029.GA15820@avx2
Signed-off-by: Alexey Dobriyan 
Reviewed-by: Andrew Morton 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm: memcontrol: expose THP events on a per-memcg basis

2019-03-06T05:07:19+00:00

Currently THP allocation events data is fairly opaque, since you can
only get it system-wide.  This patch makes it easier to reason about
transparent hugepage behaviour on a per-memcg basis.

For anonymous THP-backed pages, we already have MEMCG_RSS_HUGE in v1,
which is used for v1's rss_huge [sic].  This is reused here as it's
fairly involved to untangle NR_ANON_THPS right now to make it per-memcg,
since right now some of this is delegated to rmap before we have any
memcg actually assigned to the page.  It's a good idea to rework that,
but let's leave untangling THP allocation for a future patch.

[akpm@linux-foundation.org: fix build]
[chris@chrisdown.name: fix memcontrol build when THP is disabled]
  Link: http://lkml.kernel.org/r/20190131160802.GA5777@chrisdown.name
Link: http://lkml.kernel.org/r/20190129205852.GA7310@chrisdown.name
Signed-off-by: Chris Down 
Acked-by: Johannes Weiner 
Cc: Tejun Heo 
Cc: Roman Gushchin 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

memcg: killed threads should not invoke memcg OOM killer

2019-03-06T05:07:18+00:00

If a memory cgroup contains a single process with many threads
(including different process group sharing the mm) then it is possible
to trigger a race when the oom killer complains that there are no oom
elible tasks and complain into the log which is both annoying and
confusing because there is no actual problem.  The race looks as
follows:

P1				oom_reaper		P2
try_charge						try_charge
  mem_cgroup_out_of_memory
    mutex_lock(oom_lock)
      out_of_memory
        oom_kill_process(P1,P2)
         wake_oom_reaper
    mutex_unlock(oom_lock)
    				oom_reap_task
							  mutex_lock(oom_lock)
							    select_bad_process # no victim

The problem is more visible with many threads.

Fix this by checking for fatal_signal_pending from
mem_cgroup_out_of_memory when the oom_lock is already held.

The oom bypass is safe because we do the same early in the try_charge
path already.  The situation migh have changed in the mean time.  It
should be safe to check for fatal_signal_pending and tsk_is_oom_victim
but for a better code readability abstract the current charge bypass
condition into should_force_charge and reuse it from that path.  "

Link: http://lkml.kernel.org/r/01370f70-e1f6-ebe4-b95e-0df21a0bc15e@i-love.sakura.ne.jp
Signed-off-by: Tetsuo Handa 
Acked-by: Michal Hocko 
Acked-by: Johannes Weiner 
Cc: David Rientjes 
Cc: Kirill Tkhai 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm, memcg: extract memcg maxable seq_file logic to seq_show_memcg_tunable

2019-03-06T05:07:17+00:00

memcg has a significant number of files exposed to kernfs where their
value is either exposed directly or is "max" in the case of
PAGE_COUNTER_MAX.

This patch makes this generic by providing a single function to do this
work.  In combination with the previous patch adding
mem_cgroup_from_seq, this makes all of the seq_show feeder functions
significantly more simple.

Link: http://lkml.kernel.org/r/20190124194100.GA31425@chrisdown.name
Signed-off-by: Chris Down 
Acked-by: Johannes Weiner 
Acked-by: Michal Hocko 
Cc: Tejun Heo 
Cc: Roman Gushchin 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm, memcg: create mem_cgroup_from_seq

2019-03-06T05:07:17+00:00

This is the start of a series of patches similar to my earlier
DEFINE_MEMCG_MAX_OR_VAL work, but with less Macro Magic(tm).

There are a bunch of places we go from seq_file to mem_cgroup, which
currently requires manually getting the css, then getting the mem_cgroup
from the css.  It's in enough places now that having mem_cgroup_from_seq
makes sense (and also makes the next patch a bit nicer).

Link: http://lkml.kernel.org/r/20190124194050.GA31341@chrisdown.name
Signed-off-by: Chris Down 
Acked-by: Johannes Weiner 
Acked-by: Michal Hocko 
Cc: Tejun Heo 
Cc: Roman Gushchin 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm/memcontrol.c: use struct_size() in kmalloc()

2019-03-06T05:07:15+00:00

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array.  For example:

  struct foo {
      int stuff;
      void *entry[];
  };

  instance = kmalloc(sizeof(struct foo) + sizeof(void *) * count, GFP_KERNEL);

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

  instance = kmalloc(struct_size(instance, entry, count), GFP_KERNEL);

This code was detected with the help of Coccinelle.

Link: http://lkml.kernel.org/r/20190104183726.GA6374@embeddedor
Signed-off-by: Gustavo A. R. Silva 
Acked-by: Michal Hocko 
Cc: Johannes Weiner 
Cc: Vladimir Davydov 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

memcg: localize memcg_kmem_enabled() check

2019-03-06T05:07:15+00:00

Move the memcg_kmem_enabled() checks into memcg kmem charge/uncharge
functions, so, the users don't have to explicitly check that condition.

This is purely code cleanup patch without any functional change.  Only
the order of checks in memcg_charge_slab() can potentially be changed
but the functionally it will be same.  This should not matter as
memcg_charge_slab() is not in the hot path.

Link: http://lkml.kernel.org/r/20190103161203.162375-1-shakeelb@google.com
Signed-off-by: Shakeel Butt 
Acked-by: Michal Hocko 
Cc: Johannes Weiner 
Cc: Vladimir Davydov 
Cc: Roman Gushchin 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds